Sdílet prostřednictvím


Language - Question Answering

The Language - Question Answering API is a cloud-based service that provides advanced natural language processing with state of the art transformer models to generate answers against custom question answering projects or raw text. Question answering is a feature in Language service which is enhanced from the QnA Maker service with additional features like support for unstructured documents and precise answering.

This connector is available in the following products and regions:

Service Class Regions
Logic Apps Standard All Logic Apps regions except the following:
     -   Azure China regions
Power Automate Premium All Power Automate regions except the following:
     -   China Cloud operated by 21Vianet
Power Apps Premium All Power Apps regions except the following:
     -   China Cloud operated by 21Vianet
Contact
Name Question Answering Support
URL https://portal.azure.com/#blade/Microsoft_Azure_Support/HelpAndSupportBlade/overview
Email qnamakersupport@microsoft.com
Connector Metadata
Publisher Microsot
Website https://azure.microsoft.com/en-us/services/cognitive-services/question-answering/#overview
Privacy policy https://privacy.microsoft.com/en-us/privacystatement
Categories AI;Website

Language - Question Answering Connector is a connector to integrate any Power app with our new Question Answering feature within Language service. With this connector, users can now connect to the new as well as enhanced features of Question Answering. There are actions in this connector which help users to

  • Generate answers from project
  • Get project metadata
  • Generate answers from provided text over a period of time, we plan to roll out more actions to give more functionalities to our customers.

Prerequisites

Follow instructions here to create a Custom question answering project.

How to get credentials

  1. After deploying your project, click on “Get Prediction URL”.
  2. Get your Site URL from the hostname of Prediction URL and your Account key which would be the Ocp-Apim-Subscription-Key.PredictionURL
  3. Create a connection using your Account Key and Site URL.Connection

Get started with your connector

Use Connector in a flow

  1. Go to Microsoft Power Automate , and click on Create and select “Instant Cloud flow”.CreateFlow
  2. For testing this connector, you can click on “Manually trigger a flow” and click on Create.SelectTriggerForFlow
  3. Click on “New Step” and search “Language – Question Answering” and choose amongst the three actions.SelectActionForFlow

Actions:

Generate Answer from Project

  1. This action helps in answering the specified question using your knowledge base in your project.
  2. You can give this query for generating answer from your “Sample-project”GenerateAnswerFromProject
  3. Learn more about the request body from here: Question Answering - Get Answers - REST API (Azure Cognitive Services) | Microsoft Docs

Get Project Metadata

  1. This action helps in getting all the metadata of your project.
  2. Give your project name to get metadata of your project.GetProjectMetadata

Generate answer from provided text

  1. This action helps in answering the specified question using the provided text. To use only this action, Custom Question Answering need not be enabled on Language resource.
  2. This action doesn’t require a Custom Question Answering Project. It only needs a Language resource. You can create a language resource and copy your endpoint and key from Azure portal, in Keys and Endpoint section, to Site URL and Account Key respectively.
  3. Query “ how long it takes to charge surface? ” on the following text: “ Power and charging. It takes two to four hours to charge the Surface Pro 4 battery fully from an empty state. It can take longer if you’re using your Surface for power-intensive activities like gaming or video streaming while you’re charging it ”GenerateAnswerFromProvidedText
  4. Learn more about the request body from here: Question Answering - Get Answers From Text - REST API (Azure Cognitive Services) | Microsoft Docs

Response

Response

Generate Answer from Project

It will return an answer from your “Sample-project” for the question “Accessibility”. To know more about the response object for this action, visit Question Answering - Get Answers - REST API (Azure Cognitive Services) | Microsoft Docs

Generate Answer from provided text

It will extract answer for your question from the given text. To know more about the response object, visit Question Answering - Get Answers From Text - REST API (Azure Cognitive Services) | Microsoft Docs .

Get Project Metadata

It gets the requested project’s metadata. To know more about the response object, visit Question Answering Projects - Get Project Details - REST API (Azure Cognitive Services) | Microsoft Docs .

Known issues and limitations

We have not found any, please reach out to us if you find any at qnamakersupport@microsoft.com

Common errors and remedies

Please refer to this documentation : https://docs.microsoft.com/en-us/rest/api/cognitiveservices/questionanswering/question-answering

FAQ

Please refer to this documentation : https://docs.microsoft.com/en-us/azure/cognitive-services/language-service/question-answering/overview

Creating a connection

The connector supports the following authentication types:

Api Key ApiKey All regions Shareable
Azure AD Integrated Use Azure Active Directory to access your Azure Blob Storage. All regions Not shareable
Default [DEPRECATED] This option is only for older connections without an explicit authentication type, and is only provided for backward compatibility. All regions Not shareable

Api Key

Auth ID: keyBasedAuth

Applicable: All regions

ApiKey

This is shareable connection. If the power app is shared with another user, connection is shared as well. For more information, please see the Connectors overview for canvas apps - Power Apps | Microsoft Docs

Name Type Description Required
Account Key securestring The endpoint key for your language resource to be copied from Azure portal True
Site URL string Hostname of your language resource to be copied from Azure portal True

Azure AD Integrated

Auth ID: tokenBasedAuth

Applicable: All regions

Use Azure Active Directory to access your Azure Blob Storage.

This is not shareable connection. If the power app is shared with another user, another user will be prompted to create new connection explicitly.

Name Type Description Required
Site URL string Hostname of your language resource to be copied from Azure portal True

Default [DEPRECATED]

Applicable: All regions

This option is only for older connections without an explicit authentication type, and is only provided for backward compatibility.

This is not shareable connection. If the power app is shared with another user, another user will be prompted to create new connection explicitly.

Name Type Description Required
Account Key securestring The endpoint key for your language resource to be copied from Azure portal True
Site URL string Hostname of your language resource to be copied from Azure portal True

Throttling Limits

Name Calls Renewal Period
API calls per connection 100 60 seconds

Actions

Generate answer from Project

This action helps in answering the specified question using your knowledge base in your project.

Generate answer from provided text

This action helps in answering the specified question using the provided text.

Get Project Metadata

This action helps in getting all the metadata of your project.

Generate answer from Project

This action helps in answering the specified question using your knowledge base in your project.

Parameters

Name Key Required Type Description
Project Name
projectName True string

This is the Project name you want to use.

Question
question True string

User question to query against the project.

User ID
userId string

Unique identifier for the user.

Top #
top integer

Maximum number of answers to be returned for the question.

Confidence Threshold
confidenceScoreThreshold float

Minimum threshold score for answers. Value ranges from 0 to 1.

Ranker type
rankerType string

Type of ranker to be used.

QnA ID
qnaId string

This is the exact QnA ID to fetch from project.

Previous QnA ID
previousQnaId integer

Previous turn top answer result QnA ID.

Previous User Query
previousUserQuery string

Previous user query.

key
key string

Metadata Key from Metadata dictionary used in the QnA.

value
value string

Metadata Value from Metadata dictionary used in the QnA.

Join Metadata with
logicalOperation string

Set 'OR' or 'AND' for corresponding logical operation for metadata filters.

Source Filter
sourceFilter string

Find QnAs that are associated with any of the given list of sources in project.

Join Filters with
logicalOperation string

Logical operation used to join metadata filter with source filter.

Enable precise answer
enable boolean

Enable or disable Answer Span prediction.

Confidence Threshold
confidenceScoreThreshold float

Minimum threshold score required to include a precise answer, value ranges from 0 to 1.

Top #
topAnswersWithSpan integer

Number of Top answers to be considered for precise answer prediction from 1 to 10.

Include Unstructured Sources
includeUnstructuredSources boolean

Flag to enable query over unstructured sources.

Deployment Name
deploymentName True string

This is the specific deployment of the project you want to use.

API Version
api-version True string

This is the Client API version.

Returns

Represents List of Question Answers.

Generate answer from provided text

This action helps in answering the specified question using the provided text.

Parameters

Name Key Required Type Description
Question
question True string

User question to query against the given text records.

id
id True string

Unique identifier for the text record.

text
text True string

Text contents of the record.

Language
language string

Language of the text records. This is BCP-47 representation of a language.

String Index Type
stringIndexType string

Specifies the method used to interpret string offsets. Defaults to Text Elements (Graphemes) according to Unicode v8.0.0. For additional information see https://aka.ms/text-analytics-offsets.

API Version
api-version True string

This is the Client API version.

Returns

Represents the answer results.

Get Project Metadata

This action helps in getting all the metadata of your project.

Parameters

Name Key Required Type Description
Project Name
ProjectNamePathParameter True string

The name of the project to use.

API Version
api-version True string

This is the Client API version.

Returns

Represents the project.

Definitions

ProjectMetadata

Represents the project.

Name Path Type Description
projectName
projectName string

Name of the project.

description
description string

Description of the project.

Language
language Language

Language of the text records. This is BCP-47 representation of a language.

defaultAnswer
settings.defaultAnswer string

Default Answer response when no good match is found in the Knowledgebase.

multilingualResource
multilingualResource boolean

Resource enabled for multiple languages across projects or not.

createdDateTime
createdDateTime date-time

Project creation date-time.

lastModifiedDateTime
lastModifiedDateTime date-time

Represents the project last modified date-time.

lastDeployedDateTime
lastDeployedDateTime date-time

Represents the project last deployment date-time.

Language

Language of the text records. This is BCP-47 representation of a language.

Language of the text records. This is BCP-47 representation of a language.

Language
string

AnswersResult

Represents List of Question Answers.

Name Path Type Description
answers
answers array of KnowledgeBaseAnswer

Represents Answer Result list.

isUserQuestionAmbiguous
isUserQuestionAmbiguous boolean

Returns a boolean value if user's question was ambiguous.

KnowledgeBaseAnswer

Represents project answer.

Name Path Type Description
questions
questions array of string

List of questions associated with the answer.

answer
answer string

Answer text.

Confidence Threshold
confidenceScore float

Answer confidence score, value ranges from 0 to 1.

QnA ID
id integer

ID of the QnA result.

source
source string

Source of QnA result.

metadata
metadata object

Metadata associated with the answer, useful to categorize or filter question answers.

dialog
dialog KnowledgeBaseAnswerDialog

Dialog associated with Answer.

answerSpan
answerSpan AnswerSpan

Answer span object of QnA.

KnowledgeBaseAnswerDialog

Dialog associated with Answer.

Name Path Type Description
isContextOnly
isContextOnly boolean

To mark if a prompt is relevant only with a previous question or not. If true, do not include this QnA as search result for queries without context; otherwise, if false, ignores context and includes this QnA in search result.

prompts
prompts array of KnowledgeBaseAnswerPrompt

List of prompts associated with the answer.

KnowledgeBaseAnswerPrompt

Prompt for an answer.

Name Path Type Description
displayOrder
displayOrder integer

Index of the prompt - used in ordering of the prompts.

qnaId
qnaId integer

QnA ID corresponding to the prompt.

displayText
displayText string

Text displayed to represent a follow up question prompt.

AnswerSpan

Answer span object of QnA.

Name Path Type Description
text
text string

Predicted text of precise answer.

Confidence Threshold
confidenceScore float

Predicted score of precise answer, value ranges from 0 to 1.

offset
offset integer

The Precise answer offset from the start of answer.

length
length integer

The length of the precise answer.

AnswersFromTextResult

Represents the answer results.

Name Path Type Description
answers
answers array of TextAnswer

Represents the answer results.

TextAnswer

Represents answer result.

Name Path Type Description
answer
answer string

Answer Text.

Confidence Threshold
confidenceScore float

Answer confidence score, value ranges from 0 to 1.

id
id string

Record ID.

answerSpan
answerSpan AnswerSpan

Answer span object of QnA.

offset
offset integer

The sentence offset from the start of the document.

length
length integer

The length of the sentence.